A Captive Dashboard Cannot Make Fronting Obligations Disappear

A captive dashboard is a very agreeable object. It shows premium, paid loss, case reserves, IBNR, expected loss, capital, and a tidy retained result. It makes a complicated programme look like a portfolio. That view is useful until it becomes the view.

The captive may own the economics of the risk without owning the regulatory permission to issue the policy.

Where compulsory cover, a lender certificate, or a provincial rule calls for admitted paper, a licensed fronting carrier issues the policy and cedes the risk to our captive. So the captive is not the whole programme. It is one obligor in a chain. That changes what we choose to model. The question isn’t whether we can calculate retained loss. It is whether we can say who issued the policy. Who owes the first reimbursement. Who holds the security. Who reinsures the captive. And what fails when one can’t perform. The cleanest possible display of the captive’s economics can still conceal the front’s regulatory exposure. That is not a presentation flaw. It is a control failure with good typography.

A captive is not one kind of company

We make the first mistake before designing anything. We use “captive” as though it named one thing. It does not. A pure captive insures its owner and affiliates. An association captive writes for a trade group. A rental captive writes cells for unrelated participants. That segmentation is not decorative.

Guernsey introduced protected-cell legislation in 1997, and it separates a cell’s assets from the core and from other cells by force of law.

An incorporated cell has its own legal personality. The licence class sets the structure. The structure sets our control boundary.

Adding a line, an insured entity, or third-party business needs prior consent under the business plan.

A dashboard calling all of this “captive exposure” is numerically correct. It is also legally useless. Domicile is not a dropdown beside the name.

British Columbia has licensed captives under dedicated legislation for decades, and Alberta’s took effect in 2022.

Each regime states its own minimum paid-in capital, and the floors differ by class and by domicile. Those figures identify the regime. They are a poor definition of capital adequacy.

The binding constraint is usually the solvency margin our premium and reserves generate.

Or the front’s collateral requirement.

We quote minimum capital because it fits a slide.

It rarely stops a programme.

Location has operational consequences too.

Captives file audited statements and reserve opinions annually.

Offshore economic-substance legislation enacted from 2019 requires local core income-generating activity: adequate people, premises, and expenditure.

Board meetings. Resident management. Real underwriting decisions.

A record showing an offshore licence and none of those activities is not a view of the programme.

It makes the absence look official.

The policy follows the issuer, not the dashboard

Fronting is compelled by the direct-insurance rule.

Nobody chooses it for the fee.

A captive licensed in Barbados, Bermuda, or Vermont is non-admitted wherever it holds no direct-insurance licence.

So a front issues the policy on admitted paper, receives the premium, and cedes the risk under a reinsurance agreement.

The claim then follows that paper.

The front pays the claimant, being the issuer.

It then bills our captive.

The policyholder generally has no contractual relationship with the captive, or with the captive’s reinsurer.

A cut-through clause is a narrow exception.

Which matters as soon as our vocabulary starts implying the captive “pays claims.”

It matters more when the claim gets larger.

The front takes balance-sheet credit for a cession to an unregistered reinsurer only so far as the obligation is secured.

A clean irrevocable letter of credit. A qualifying reinsurance trust. Funds withheld.

In Canada, OSFI grants credit for unregistered reinsurance against collateral held under a reinsurance security agreement.

Guideline B-3 adds a board-approved reinsurance plan, counterparty diligence, and limits on ceding substantially all risk.

The programme can also use funds withheld, with the front retaining ceded premium rather than remitting it.

That does not make collateral disappear.

It changes where the security sits.

When the front’s estimate of ultimate ceded liabilities exceeds available security, it can call for more.

So a higher IBNR estimate creates a cash demand.

Our dashboard still shows a satisfactory loss ratio.

“The dashboard says the captive is profitable” will not move a front whose regulator has withdrawn balance-sheet credit.

Reinsurance has an order, not just a total

The captive buys its own protection, and we must not flatten it into one recoverables number.

Quota share takes a fixed percentage of every premium and loss dollar in a covered class, from the first dollar, usually against a ceding commission.

Surplus share is proportional too.

Its ceded percentage varies risk by risk with the retained line and the number of lines.

Excess of loss works differently.

It responds above an attachment, to a limit.

Those distinctions turn operational when a claim arrives.

The treaty decides what inures first.

A catastrophe excess layer written net of quota share attaches to the loss left after the proportional cession.

Written gross, the tower sees a different loss.

No calculation corrects a misread inuring clause.

Nor is every recovery an asset because a model says so.

A contract earns reinsurance accounting only when it transfers significant insurance risk, underwriting and timing both, with a real possibility that the reinsurer loses money.

The familiar 10-10 screen is market convention, not an accounting standard: roughly a 10% probability of a 10% present-value loss.

A failed test brings deposit accounting.

No reserve credit. No earnings benefit.

IFRS 17 keeps reinsurance held separate from the underlying business.

We cannot simply net them.

So we preserve the underlying obligation, the reinsurance held, the loss-recovery component for an onerous group, and the terms behind the accounting conclusion.

One net retention tile is not an accounting model.

The chain of security is the management view

I would build the view on events, not charts.

A policy issuance names the admitted issuer, the jurisdiction, the insured, the policy term, and the captive reinsurance agreement.

A collateral event names funds withheld, letter of credit, or trust; the current amount; the estimate of ultimate liabilities; and the date and reason for any call.

A claim event records who paid the claimant, the front’s bill to the captive, the case reserve, the IBNR, and the effect on collateral.

Then comes the treaty record.

It needs treaty type, counterparty, signed percentage, attachment, limit, inuring order, authorization status, and reporting mechanics.

Proportional treaties settle through quarterly bordereaux and a statement of account netting ceded premium, brokerage, ceding commission, and paid losses.

Excess-of-loss recoveries settle against loss advices.

These are programme events.

They are not attachments to a finance workflow.

Our counterparty is not one row called “panel.”

Subscription participants are severally, not jointly, liable.

Each answers for its own signed share.

If one fails, the others don’t fill the gap.

So we need the participant-level line, the rating and concentration, the security status where one is required, and an overdue-reporting exception.

We need to know when a cash call is available.

That clause lets a cedent demand payment before it has finished paying claimants.

This is board material.

It is not a board dashboard.

Five questions, in order, are what I would want to answer before walking into that room.

Who issued.

Who retained.

Who reinsures.

Who holds collateral.

And what happens when one can’t perform.

The first four describe the design.

The fifth tells us whether we understand it.

Canada is not a branch setting

A US playbook gives us language and the wrong controls.

The NAIC model-law amendments of 2019 reduced collateral to zero for qualifying reciprocal-jurisdiction reinsurers.

Virtually no captive qualifies.

Not a generic permission to erase collateral from a Canadian fronted programme.

Tax is the other boundary.

It is the one we cross by accident.

A captive licence says a regulator has permitted the captive to operate under its capital and business-plan rules.

It says nothing about deducting our premium.

Canadian foreign-accrual-property-income rules can treat a foreign affiliate’s income from insuring Canadian risk as passive, and taxable here now.

Anti-avoidance amendments enacted in 2014 addressed swap arrangements designed to turn Canadian risk into foreign risk.

OECD guidance issued in 2020 addresses captives directly.

Genuine diversification, capital at risk, real underwriting decisions: those are the marks of an insurer.

The useful record carries domicile, licence class, business plan, capital and reserve basis, substance evidence, tax decision records, and the contracts that actually move the risk.

It reaches no tax conclusion.

It makes ours reviewable.

Here is the part nobody wants written down.

Don’t buy a dashboard that feels self-contained.

Build controls that make its dependencies hard to ignore.

A captive is most useful when its dashboard shows the obligations it cannot retain for itself.